Training the ANN

To start training an ANN, all weights and biases associated with the individual input (in other words, how important each input is to correctly predicting the observed data) are initially randomly assigned values. This way, the ANN produces some output data that in the first step has no relationship with the observed data. Then the output of every output neuron gets compared with the expected data and an error is calculated for every output neuron. The error is then back-propagated to the neurons in the hidden layer, weighted by the respective factors. For the hidden layer, an error for every neuron is calculated as well. After this the weights are changed according to their value and the error. For each data pair to be learned a forward pass and backwards pass is performed. This is repeated over and over again until the error is at a low enough level. By applying this technique repeatedly the ANN adapts to the given problem, i.e., for the given inputs it produces outputs close to those expected.

Weights are changed by an amount proportional to the error at that unit times the output of the unit feeding into the weight. Running the network consists of

Forward pass: the outputs are calculated and the error at the output units calculated. Backward pass: The output unit error is used to alter weights on the output units. Then the error at the hidden nodes is calculated (by back-propagating the error at the output units through the weights), and the weights on the hidden nodes altered using these values. For each data pair to be learned a forward pass and backwards pass is performed. This is repeated over and over again until the error is at a low enough level (or we give up).

Back to Neural Network Application to Water Level Forecasting